Jiaqi Ma | 马家祺

Biography

I’m interested in the broad area of trustworthy artificial intelligence (AI). We recognize that AI models often live in complex ecosystems, where ensuring trustworthiness involves dealing with challenges from heterogeneous data and intricate human interactions, and adhering to a variety of public policies. My research blends observations and insights of these ecosystem dynamics into innovative technical solutions, aiming to develop trustworthy AI systems that reliably operate within their specific ecosystems. Specifically, my work addresses the following major questions with some detailed examples:

1. How to effectively operationalize regulatory principles such as explainability, fairness, privacy, and address legal considerations such as copyright?
2. Are there inherent technical conflicts or trade-offs involved in simultaneously enforcing multiple regulatory principles?
3. How can we provide technical solutions to assist policymakers in addressing copyright issues of generative AI?
4. How to leverage the knowledge about human interactions with the AI systems to better design the system?
5. How can we effectively leverage multiple user signals to improve user satisfactory in recommender systems?
6. How can we better accommodate downstream needs of domain scientists in drug discovery?
7. How to deal with challenges for learning from complex real-world data?
8. How to characterize and improve fairness and robustness for learning from graph-structured data?
9. How to learn from partially observed data, such as partial rankings or censored survial data?

Some “buzzwords” relevant to my existing research include trustworthy machine learning, explainable machine learning, machine unlearning, graph machine learning, recommender systems, and large language models.

Selected PhDs

Anyone you know?

Selected Masters

Anyone you know?

Recruitment Information

For students who want to work with me, please see here for more details.

Comments (optional)

Share working/internship experience

• start-date - end-date : xxxxx
• .....